Study Design and Participants: This single-center, retrospective case-control study enrolled adult patients (18-100 years) hospitalized in the RICU of the Chinese PLA General Hospital between January 2018 and December 2024, who were clinically diagnosed with pulmonary infection and underwent both bronchoalveolar lavage (BAL) and rapid on-site evaluation (ROSE) examination. Exclusion criteria included age outside the range, non-infectious pulmonary diseases, absence of bronchoscopy/BAL, unqualified ROSE specimens (poor image quality or substandard samples), or missing critical clinical data for grouping.

Diagnostic Criteria and Grouping: Aspiration pneumonia (AP) was diagnosed using a clinical-imaging composite gold standard. The AP group had to simultaneously meet: (1) chest CT showing new gravitationally-dependent infiltrates; (2) a definite aspiration history or at least one high-risk factor (e.g., impaired consciousness, swallowing dysfunction, gastroesophageal reflux, diminished cough/gag reflex, poor oral hygiene); and (3) ≥2 of the following infection criteria: fever (>38.0°C or <36.0°C), new/worsened respiratory symptoms or purulent secretions with gastric contents, and abnormal peripheral white blood cell count or left shift. The non-AP group comprised patients with ≥2 infection criteria but who did not meet the AP diagnostic criteria.

ROSE Procedure and Definition of "Showering Cells": BALF was submitted for testing within 30 minutes, centrifuged, and the sediment smeared onto slides for Gram and Diff-staining. A blinded morphologist assessed specimen adequacy and performed microscopic examination. "Showering cells" were defined as squamous epithelial cells observed under high-power fields (×1000) with the distinct morphology of no bacterial adhesion on the surface, clear cytoplasm, and a clean background. Ten non-overlapping high-power fields were randomly selected to count the total number of showering cells and total squamous epithelial cells, and the showering cell ratio (showering cells / total squamous epithelial cells) was calculated.

Data Collection: Demographic data, aspiration risk factors, imaging features, inflammatory markers (white blood count, neutrophil percentage, C-reactive protein, Interleukin-6,Procalcitonin), showering cell-related indicators, underlying diseases, and respiratory/nutritional support modalities were collected.

Statistical Methods: Continuous variables were expressed as median (IQR) and compared using the Mann-Whitney U test; categorical variables were compared using the χ² or Fisher's exact test. For multivariate analysis, multiple imputation and complete case analysis were jointly used to construct logistic regression models (controlling for confounders, VIF<5), calculating adjusted odds ratios (aOR) and 95% CIs, with the showering cell ratio scaled by a factor of 10. Diagnostic performance was evaluated via ROC curves (AUC), with the optimal cutoff determined by the Youden index; AUCs were compared using the DeLong test, and internal validation was performed using the Bootstrap method (1000 resamples). For incremental value assessment, a baseline model (core clinical variables) and a combined model (adding showering cell indicators) were constructed. Discriminative ability was compared via AUC, calibration was assessed by calibration curves and mean absolute error, and reclassification improvement was quantified using NRI and IDI. Decision curve analysis (DCA) was applied to evaluate clinical net benefit, and a nomogram was ultimately constructed based on the optimal model. A two-sided P <0.05 was considered statistically significant.

Inclusion Criteria:

1. Age 18-100 years;
2. Clinically confirmed pulmonary infection;
3. Completion of bedside BAL;
4. Qualified BALF specimens submitted for M-ROSE examination;
5. Complete clinical and radiological data.

Exclusion Criteria:

1. Age <18 years or >100 years;
2. Diagnosis of noninfectious pulmonary diseases;
3. No bronchoscopy or BAL examination;
4. BALF samples not assessed by M-ROSE, acquired images were of poor quality, or samples were unqualified;
5. Missing key clinical data precluded group assignment.

Inpatients with confirmed lung infection admitted to Chinese PLA General Hospital.